Semiconductor integrated circuits (ICs) are generally made by a sequence of steps including one or more exposures of a photoresist to light through a patterned mask. The diffraction of light imposes limits on the fineness of detail that can be produced by exposures of this kind, and as a result, the density of devices that can be manufactured on a single substrate is limited, in part, by the choice of exposing wavelength. In order to increase the device density, practitioners of IC manufacture have begun developing techniques involving the exposure of special resists to electromagnetic radiation of extremely short wavelengths, such as ultraviolet radiation and x radiation.
In one approach, referred to as proximity print x-ray lithography (PPXRL), hard x rays, having wavelengths of 0.3-5 nm, expose a substrate through a pattern of x-ray absorbing material (such as gold or tungsten) supported on a membrane which is transmissive to the x rays. This method can produce linewidths as small as 20 nm. However, PPXRL has posed significant technical difficulties. In particular, it has been difficult to provide an x-ray source of the required brightness, it has proven difficult to manufacture the masks, and the supporting membranes are generally somewhat fragile.
A second approach to x-ray lithography, referred to as soft x-ray projection lithography (SXPL), is described, e.g., in U.S. Pat. No. 5,003,567, issued on Mar. 26, 1991 to A. M. Hawryluk et al. This approach takes advantage of recent advances in the field of x-ray optics. For example, it is now possible to build an x-ray reduction camera using curved imaging mirrors. These mirrors may be spherical or aspheric. Each mirror includes a substrate of a material such as glass-ceramic or sintered glass having a low coefficient of thermal expansion. The first surface of the substrate is typically ground to high precision and polished. This surface is then overcoated with a multilayer coating, typically a periodic multilayer of material pairs (although groups of more than two alternating materials may be used). The alternating (e.g., paired) materials have a large difference in complex index of refraction at the x-ray wavelength being used. As a consequence of the periodic variation of complex refractive index, the mirror exhibits high x-ray reflectivity at certain angles of incidence. A typical x-ray reduction camera uses a reflective mask consisting of a thin, IC metallization pattern overlying an x-ray-reflective, multilayer coating on a polished (flat or curved) substrate surface. The mask is positioned such that x rays incident thereupon are reflected from the mask onto a primary mirror, from there onto one or more secondary mirrors, and from the last secondary mirror onto a wafer surface coated with an appropriate resist. Image reductions as great as 20:1 have been achieved in this way. (See e.g., D. W. Berreman et al., Opt. Lett. 15(1990) 529-531.)
The most promising multilayer coated optical elements (i.e., mirrors and masks) include coatings based on metal-silicon bilayers, in which the metal is, for example, molybdenum, rhodium, or ruthenium. These coatings are suitable for use at x-ray wavelengths of 130 .ANG.-300 .ANG., which, in energy, lie below the silicon L-edge near 125 .ANG. and consequently are relatively weakly attenuated by the silicon layers. For use at even shorter wavelengths, multilayer coatings can be designed to take advantage of the low absorption of other elements such as beryllium, boron, and carbon.
Metal-silicon multilayer coatings are typically deposited by DC magnetron sputtering in argon. For molybdenum-silicon coatings, the total number of bilayers deposited typically ranges from 20 up to about 60, and the bilayer spacing typically ranges from about 68 .ANG. to about 75 .ANG..
The economic importance of maintaining highly reflective optical elements is discussed in N. M. Ceglio, et al., "Soft X-Ray Projection Lithography System Design", OSA Proceedings on Soft-X-Ray Projection Lithography, 1991, Vol. 12, J. Bokor, ed., Optical Society of America (1991) 5-10. As explained therein, the exposure-limited throughput of a SXPL manufacturing system is very strongly dependent on the mirror reflectivity. Indeed, a decrease of mirror reflectivity from 70% to 50% could theoretically increase the cost of manufactured wafers by 1000%. However, the reflectivity of multilayer optical elements is expected to decrease over time as a result of environmental damage and aging effects. In order to maintain an adequate throughput, operators of a manufacturing system will have to replace or repair degraded optical elements.
In addition, it may be necessary to strip, i.e., remove, multilayer coatings during or immediately after the original fabrication procedure if, for example, the multilayer coatings have poor morphology, causing low reflectance, or if they have high reflectance but at the wrong wavelength.
Expected replacement costs are very high. This point is discussed, for example, in D. P. Gaines, et al., "Repair of high performance multilayer coatings", SPIE Vol. 1547 Multilayer Optics for Advanced X-Ray Applications(1991) 228-238. According to that article, the optical elements of a diffraction limited system operating at 130 .ANG. must maintain less than 10 .ANG. figure error. Moreover, in order to have high peak reflectivity, the surface roughness must generally be less than about 1 .ANG. over spatial wavelengths as short as about 100 .ANG. (for an x-ray wavelength of 140 .ANG.). Fabrication of blanks, particularly curved blanks, to these tolerances is time-consuming and expensive. As a consequence, it is economically attractive to repair optical elements rather than to replace them.
Practitioners in the field of x-ray lithography have, in fact, addressed the problem of repairing multilayer coated optical elements. For example, the above-cited article by D. P. Gaines, et al. describes two repair methods. One is a method of overcoating defective multilayer coatings with new multilayer coatings, and the other is a method of stripping the entire defective multilayer coating by etching an underlying release layer. Of these two methods, the stripping method may be more generally useful, because overcoating will not cure certain defects. These defects include increased surface roughness, departure of a mirror from its required figure, and macroscopic defects such as delamination and cracking. In such cases, the old multilayer coating must be stripped and replaced. However, the use of an underlying release layer may pose problems, because the time required to remove a multilayer coating by this technique increases rapidly with increasing surface area.
As noted, the stripping method of the above-cited article calls for a release layer to be deposited on the substrate before the reflective multilayer is deposited. To be useful in this regard, the release layer must be uniform, it must provide an extremely smooth surface for subsequent deposition thereupon of the multilayer, and it must generally be etchable in an etching solution that is relatively harmless to the substrate. (An etchant is relatively harmless if in the course of ordinary etching times it will not roughen the substrate beyond acceptable tolerances.) Gaines, et al., cited above, reports the use of an aluminum release layer. Aluminum was selected because, according to that article, it can be uniformly deposited and can be etched in, e.g., a solution of hydrochloric acid and cupric sulphate without measurable damage to the surface finish of a silicon-based substrate. However, aluminum release layers were found to reduce the peak, normal-incidence, x-ray reflectivities of overlying multilayers by a significant amount. This was attributed to surface roughness at spatial wavelengths smaller than 2.5 .mu.m.
At present, there is no assurance that any material will satisfy all of the requirements for a release layer completely enough to provide a practical method for repairing optical elements. Yet another problem with the use of a release layer is that the process of removing the multilayer coating is relatively time-consuming, as noted above. This is because in order to attack the release layer, the etchant must first penetrate the multilayer coating. Because it is generally dissolved slowly, if at all, by the etchant, the multilayer coating serves as an effective etch barrier which can delay, or even prevent, the dissolution of the release layer.
An alternative method of removing the multilayer coating is to etch it directly. However, this approach has encountered difficulties because at least three different materials are involved. That is, a direct etching process must remove both components (e.g., the metal and silicon components) of the multilayer coating, while maintaining sufficient selectivity to avoid attacking the substrate. In general, etchants able to remove both components in one step have not been found selective enough to avoid damaging the substrate. On the other hand, highly selective etchants have generally been found capable of removing only one component or the other. This makes it necessary to remove the multilayer coating in many, alternating steps, which is undesirable because it is relatively time-consuming. Thus, practitioners in the field have hitherto failed to provide a practical method for directly etching away the multilayer coating.